ecutwfc not converge with NC pseudopotentials

[yanghuang]

Dear users and developers,
For my system (semiconductor), ecutwfc must be at least 100 Ry to convergence, compare to VASP, which just need 38 Ry (PAW)
the value of ecutwfc is too large to carry out following Yambo calculation
In addition, my system include Pb, so SOC effect should be considered.

How should I resolve this problem?

Thanks in advance

Best,
Yang


scf.in
&control
calculation='scf'
restart_mode='from_scratch'
pseudo_dir = './',
outdir = './'
wfcdir = './'
prefix='a',
nstep = 200,
verbosity = 'high'
wf_collect = .true.
/
&system
ibrav = 0,
nat= 50, ntyp= 6,
ecutwfc = 38
lspinorb= .false.
noncolin= .false.
/
&electrons
conv_thr = 1.0d-8
mixing_beta = 0.7
diagonalization = 'david'
/
CELL_PARAMETERS {angstrom}
18.58 0 0
0 6.267 0
0 0 6.466
ATOMIC_SPECIES
Pb 207.2 Pb.rel-pbe-n-nc.UPF
I 126.9 I.rel-pbe-n-nc.UPF
S 32.065 S.rel-pbe-n-nc.UPF
C 12.017 C.rel-pbe-nc.UPF
N 14.006 N.rel-pbe-nc.UPF
H 1.007 H.rel-pbe-n-nc.UPF
ATOMIC_POSITIONS {crystal}
Pb 0.500000000 0.394727599 0.217545602
Pb -0.000000000 0.605272431 0.717545631
I 0.500000000 0.884561011 0.296939592
I 0.000000000 0.115438989 0.796939592
I 0.500000000 0.344483771 0.733856670
I -0.000000000 0.655516259 0.233856640
S 0.343479961 0.321804295 0.220015148
S 0.156520039 0.678195675 0.720015133
S 0.843479931 0.678195675 0.720015133
S 0.656520069 0.321804295 0.220015148
C 0.311365212 0.561857292 0.267952701
C 0.188634788 0.438142708 0.767952701
C 0.811365242 0.438142708 0.767952701
C 0.688634758 0.561857292 0.267952701
C 0.358918877 0.845233902 0.813741632
C 0.141081123 0.154766083 0.313741632
C 0.858918877 0.154766083 0.313741632
C 0.641081123 0.845233902 0.813741632
N 0.287089254 0.737337360 0.302407531
N 0.212910746 0.262662640 0.802407561
N 0.787089284 0.262662640 0.802407561
N 0.712910716 0.737337360 0.302407531
N 0.335805706 0.010666556 0.649089402
N 0.164194294 0.989333448 0.149089402
N 0.835805706 0.989333448 0.149089402
N 0.664194294 0.010666556 0.649089402
H 0.292551441 0.107328769 0.703937407
H 0.207448559 0.892671261 0.203937407
H 0.792551411 0.892671261 0.203937407
H 0.707448589 0.107328769 0.703937407
H 0.318986940 0.929479720 0.512734746
H 0.181013060 0.070520276 0.012734746
H 0.818986940 0.070520276 0.012734746
H 0.681013060 0.929479720 0.512734746
H 0.379460721 0.111293329 0.612489917
H 0.120539279 0.888706679 0.112489917
H 0.879460691 0.888706679 0.112489917
H 0.620539309 0.111293329 0.612489917
H 0.379992995 0.930697521 0.949130018
H 0.120007005 0.069302501 0.449130018
H 0.879992965 0.069302501 0.449130018
H 0.620007035 0.930697521 0.949130018
H 0.401037203 0.743481616 0.746125985
H 0.098962797 0.256518369 0.246125985
H 0.901037233 0.256518369 0.246125985
H 0.598962767 0.743481616 0.746125985
H 0.311406405 0.750091955 0.855379639
H 0.188593595 0.249908060 0.355379639
H 0.811406405 0.249908060 0.355379639
H 0.688593595 0.750091955 0.855379639
K_POINTS {automatic}
1 4 4 0 0 0

[Daniele Varsano]

Dear Yang Huang,
The fact you need 100Ry to converge the ground state does not mean that you need all the plane waves for the GW/optics calculations. In general you need much less, even if a large cutoff for ground state would probably imply a big number of G vectors also for Yambo.
I do not know what kind of pseudo you are using anyway recent NC pseudopotentials by Don Hamann (ONCVPSP) are very accurate and softer than the usual one. You can download the code and documentation to generate them in UPF format here:
http://www.mat-simresearch.com/

They are also collected here:
http://www.pseudo-dojo.org/
and in a dedicated github page, but I do not know if in the table (git project) the UPF format needed by QE is supported.

Best,

Daniele

[yanghuang]

Dear Daniele,
Thank you so much for your kind help!

Best,

Yang